Essays on fiscal sustainability and tax smoothing and fiscal policy simulation experiments for sri lanka

SUMMARY The focus of this thesis is on the following issues related to fiscal policy: sustainability of fiscal deficits, validity of the tax smoothing hypothesis, and macroeconomic impac

ESSAYS ON FISCAL SUSTAINABILITY AND TAX SMOOTHING, AND FISCAL POLICY SIMULATION EXPERIMENTS FOR SRI LANKA

J M ANANDA JAYAWICKRAMA

NATIONAL UNIVERSITY OF SINGAPORE

2006

ESSAYS ON FISCAL SUSTAINABILITY AND TAX SMOOTHING, AND FISCAL POLICY SIMULATION EXPERIMENTS FOR SRI LANKA

J M ANANDA JAYAWICKRAMA

[MA (Thammasat)]

[BA (Hons.) (Peradeniya)]

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF ECONOMICS NATIONAL UNIVERSITY OF SINGAPORE

2006

ACKNOWLEDGEMENTS

I received assistance from many quarters on this research First, I would like to express my sincere appreciation to Associate Professor Tilak Abeysinghe for his valuable advisory contribution to this thesis I appreciate Associate Professor Habibullah Khan for guidelines offered, especially in the early stage of the research I gratefully acknowledge helpful comments and suggestions made by Professor Åke Blomqvist (Chairman, Oral Panel) and examiners Professor Alfred Haug (York University), Associate Professor Aditya Goenka and Associate Professor Chia Ngee Choon

I express my sincere gratitude to the NUS for providing me an excellent opportunity, financial and other facilities to complete a PhD degree I would appreciate the University of Peradeniya too for the support offered I would like to thank Professor W M Sirisena, Professor W M Tilakeratne, Mr S G Liyanage, Mr Palitha Pathberiaya, Dr N D Samarawickrame for their support and encouragement

On a more personal note, I express my sincere gratitude to my late father and

my mother for giving me the ability and strength to carry out this task I am really indebted to my father who toiled hard to make our lives better off It is my feeling that

my higher education was ornamental but not quite instrumental for my father’s life I thank my brothers and their families for the support and encouragement I am thankful for my relatives and friends for the support extended I am always grateful for my wife Anoma, son Menake and daughter Kaveesha for their continuing support and understanding

Ananda Jayawickrama NUS

December, 2006

CHAPTER TWO: ON THE SUSTAINBILITY OF FISCAL DEFICITS:

THE UNITED STATES EXPERIENCE

2.7 Trends in Deficit and Debt of the U.S Federal Government 27

CHAPTER THREE: TAX SMOOTHING HYPOTHESIS REVISITED:

EXPERINCE OF SOME OECD ECONOMIES

3.2 Tax Smoothing Hypothesis and Previous Studies 43

3.5 Derivation of the Permanent Expenditure Rate 61

CHAPTER FOUR: A MACROECONOMETRIC MODEL FOR SRI LANKA

FOR POLICY SIMULATION EXPERIMENTS

4.3.5 Accounting and Definitional Identities 134

CHAPTER FIVE: SIMULATION EXPERIMENTS ON GOVERNMENT

EXPENDITURE POLICIES IN SRI LANKA

5.3 Simulation Experiments on Government Expenditures 159

5.3.4 A Comparative Analysis on Expenditure Multipliers 168

6.3 Simulation Experiments on Government Spending in Sri Lanka 181

APPENDIX III: COINTEGRATION AND ERROR CORRECTION

SUMMARY

The focus of this thesis is on the following issues related to fiscal policy: sustainability of fiscal deficits, validity of the tax smoothing hypothesis, and macroeconomic impact of government expenditure policies We propose new methodological approaches to the issues of the fiscal sustainability and the tax smoothing hypothesis The fiscal sustainability and the tax smoothing hypotheses are, then, tested using fiscal data of selected developed countries As for developing countries,fiscal policy issues are not indeed fiscal sustainability or tax smoothing but how to contain fiscal deficits and effects of such deficit reduction measurers Simulation experiments are, therefore, carried out on economic effects of deficit reduction policies in a developing country context taking Sri Lanka as the case

We examine the sustainability of the U.S federal government budgetary policies by extending existing present value borrowing constraint model Using rational expectations to allow for full information in the present value borrowing constraint, the sustainability of the U.S budget deficits is examined in a longer time horizon that includes 75 years Results emerge in favour of the sustainability of the U.S federal fiscal deficits The model developed here is rich enough explaining very divergent movements in the debt series without any artificially defined structural breaks or regime shifts

We propose a new theoretical and empirical framework for the tax smoothing hypothesis For this first we derive a linear relationship between the optimal tax rate and the permanent component of the government expenditure rate Using this linear relationship between the optimal tax rate and the permanent government expenditure rate, we show that the random walk implication of the tax smoothing hypothesis is

valid if the tax rate at time t and the permanent government expenditure rate at time

t-1 are cointegrated with a vector (1 1) The general conclusion of this study

depending on the degree of the cointegration and results of an error correction model

is that countries in the sample follow aweak form of tax smoothing

In the next two chapters, we examine the macroeconomic impact of government spending policies in Sri Lanka For this, a macroeconometric model is constructed for Sri Lanka The model is simulated to trace-out the impact of decreases

in government consumption, investment and interest payment spending Simulation results reveal that while government consumption and transfer payment spending cuts leave many macro variables unchanged, government investment expenditure cuts have significant impact on them It is found that lowered government investment spending results in a severe economic recession While low governmentconsumption expenditure and transfer payments decrease fiscal deficit markedly, low governmentinvestment spending triggers a recession and results in higher fiscal deficits in subsequent years

LIST OF TABLES

Table 2.1: Fiscal Summary of the U.S Federal Government (as a % of GDP) 28 Table 2.2: Computation of Net Debt and Adjusted Primary Balance 29 Table 2.3: ADF Test Results for a Unit Root in Variables in (2.11a) 34

Table 3.1: ADF Test for a Unit Root in Expenditure Rate 65

Table 3.3: Test for Cointegration Between Tax and Permanent Expenditure Rates 71

Table 3.5: Test for Causality Between Permanent Expenditure and Tax Rates 74

Table 4.2: ADF Unit Root Test on Variables Involving Stochastic Equations 102 Table 5.1: Summary of Fiscal Policy Indicators of Sri Lanka, 1975-2004 155 Table 5.2: Impact of Temporary Cut in Government Consumption Spending 161 Table 5.3: Cumulative Effect of Government Consumption Spending Cut 162 Table 5.4: Impact of Temporary Cut in Government Investment Spending 164 Table 5.5: Cumulative Effect of Government Investment Spending Cut 165 Table 5.6: Impact of Temporary Cut in Government Transfer Payments 167 Table 5.7: Cumulative Effect of Government Transfer Payments Cut 168 Table AIV.1: Capital Stock Estimates for Sri Lanka, 1967-2004 230 Table AIV.2: Actual and Estimated Data for Employment and Unemployment 234 Table AIV.3: Weights Assigned to Major Trading Partners of Sri Lanka 238 Table AIV.4: Major Trade Partners’ Currency and Computed Trade-weighted

Exchange Rate for Sri Lanka in Domestic Currency 240 Table AIV.5: Major Trade Partners’ Income and Computed Trade-weighted

World Income (U.S Dollar Billion in 2000 Prices) 241

LIST OF FIGURES

Figure 2.1: Federal Government Fiscal Balance and Debt (U.S Dollar Billion) 30 Figure 2.2: Actual and Fitted Debt Without the Lagged Dependent Variable 35 Figure 2.3: Recursive Estimates and Actual and Fitted Net Debt 38 Figure 2.4: Impulse Response Effects of Each Variable on the Debt Stock 39 Figure 3.1: Actual, Unit Root and Smooth Series of Expenditure Rate 65

Figure 4.1: Time Series of Private Investment and Output, 1978-2004 109

Figure 4.3: Actual and Static Simulation Results of the SLM 146 Figure 4.4: Actual and Dynamic Simulation Results of the SLM 147 Figure 5.1: Fiscal Deficits and Debt Accumulation in Sri Lanka, 1975-2004 157 Figure 5.2: Impact Multipliers of Government Spending Shocks 170 Figure 5.3: Cumulative Effect of Government Expenditure Shocks 171 Figure 5.4: Change in Fiscal Deficit Ratio in Response to Spending Cuts 175 Figure AIV.1: Incremental Capital Output Ratio in Sri Lanka, 1960-2004 230 Figure AIV.2: Actual and Fitted Values of Employment, 1990-2004 233 Figure AIV.3: Actual and Fitted Values of Unemployment Rate, 1990-2004 236

LIST OF CHARTS

CHAPTER ONE INTRODUCTION 1.1 Background of the Study

The financial stress of the public sector is one of the major recent economic issues

in both developed and developing countries Over the last three decades, there has been a tendency for large fiscal deficits to appear Recurrent and high deficits result in excessive accumulation of government debts Deficits and debt accumulation raise a number of issues on government budgetary policies One issue is the sustainability of fiscal deficits, and another is whether countries follow optimal fiscal policy criteria such as tax smoothing There are also concerns on whether fiscal deficits and deficit reduction policies have real effect on the economy It is argued that in the absence of Ricardian equivalence fiscal deficits have large impact on the macroeconomy as they disturb the private sector decision-making process.1 Further, fiscal deficits cause intergenerational burdens since most of the excess expenditures are financed by issuing debt As Romer (2006, p.568) notes, budget deficits reduce growth and could lead to a crisis if they are highly persistent and too large A large volume of research papers have emerged discussing these various aspects of budgetary policies of the government

The management of the government budgetary policies has varied over time Before the Keynesian economics emerged, the fiscal conduct was mainly governed by

1 The Ricardian equivalence theorem states that only the quantity of government expenditure, not the division of financing government expenditure between taxes and bonds, affects the economy This is true if agents accurately foresee future tax liabilities implicit in deficit financing Since agents know that their future tax liabilities are exactly equal to the return from bond holding, this policy does not affect the agents’ wealth thus consumption and investment (see Romer (2006, p 568)) This view is originally proposed by David Ricardo and elaborated later by Robert Barro (see Barro (1974)) Seater (1993) provides an extensive survey of literature on the Ricardian equivalence theorem

the principle of fiscal responsibility proposed by Adam Smith From Adam Smith’s point of view, no difference is found between budgetary accounts of a family and of the government He, therefore, advises policymakers to maintain government revenue and spending accounts in balance (see Buchanan and Wagner (2000)) The government should not spend without imposing taxes It should not shift the present fiscal burdens to future generations by bond financing excess spending in favour of the current generation The policy rule was to run a balanced budget If not, the government budget should be in surpluses Deficits were tolerated only in extraordinary times such as wars or recessions Surpluses attained in normal periods should be sizable to cushion unexpected deficits In consequence, the debt stock increased only in extraordinary times (see for discussion Burkhead (1954), Buchanan (1958), Buchanan and Wagnar (1967))

The publication of Keynes’ The General Theory of Employment, Interest and

Money in 1936 had a profound influence on the fiscal policy conduct Keynes

proposed that budgetary conduct of the government is totally different from that of individuals or firms A policy measure that is folly in the conduct of a family budget may be prudence in the conduct of the budget of a nation (Keynes (1936), see also Buchanan and Wagnar (2000)) In Keynesian economics a key role is assigned to the government budget as it was used to achieve more important macroeconomic objectives such as growth and economic stabilization at full employment level (Pierce (1971)) Large deficits, thereby the accumulation of debt, are not worrisome if excess expenditures are used to enhance the production capacity and for the stabilization of the economy In addition, Keynesianism pointed out that deficits do not matter if they are financed by debts within the nation (Dalton (1954), Feldstein (1995)) These ideas

result in recurrent and large budget deficits as policymakers were no longer obliged to balance the budget

However, this unbalanced budget rule has often led policymakers to make seemingly irresponsible expenditure decisions (see Feldstein (1995)) Unconstrained spending rules paved the way for various rent-seeking activities too It is learnt that in

a non-Ricardian world, fiscal deficits drive up interest rates and crowd-out private investment and result in an erosion of the longer term productivity growth (see for recent work Mühleisen (2004), Adam and Bevan (2005)) Feldstein (1995) noted that deficits also create a massive deadweight loss to the economy

As governments understood the consequences of large fiscal deficits, recent trend has been to narrow down the gap between government spending and revenue In many countries, fiscal deficits have fallen sharply from their high figures in the 1990s (Auerbach (2003), Adam and Bevan (2005)) One reason for the tightening fiscal policy in recent years was the high accumulation of debt in the past (Bohn (1998)) Another reason, particularly in developed countries, was the rejection of Keynesian view that deficits are benign Because of its contemporaneous effects and intergenerational burdens, many prefer to reduce (or eliminate) fiscal deficits by lowering government expenditure (Feldstein (1995)) On the other hand, declining fiscal deficits in many developing countries are a result of policy targets set by international funding agencies to qualify for further financing As recent foreign funding is associated with policy packages aimed at downsizing the government and deregulating towards free market economy, deficits tend to decline However, it is evident that governments regularly create fiscal deficits aiming to garner political gains Fiscal deficits are, therefore, hard to defy and, as a result, the accumulation of

debt is inevitable To reiterate, this kind of fiscal policy obviously paves the way for many macroeconomic concerns

1.2 Statement of the Research Problems

One issue that this thesis addresses is the sustainability of the fiscal deficits in the long-run We examine this issue to assess whether the current budgetary policy would ultimately lead the government into insolvency Another issue that is taken up

in this thesis is the validity of the tax smoothing hypothesis Given that deficits are cyclical and taxation creates an excess burden, we examine whether governments follow a tax smoothing pattern in order to spread out the tax burden over time Third,

we question whether different components of government expenditure affect the aggregate economy differently

The objectives of this study are three-fold;

1 Offer a new methodology to assess fiscal sustainability

2 Offer a new test procedure for testing the tax smoothing hypothesis

3 Assess the macroeconomic impact of various deficit reduction methods We

do this exercise for Sri Lanka, a less developed country stuck in a prolonged civil war For Sri Lanka, both fiscal sustainability and tax smoothing are far from reality Chronic deficit is the major problem regarding government budgetary policies How to reduce the deficit is a main concern For this, we construct a macroeconometric model Another objective is to use this model

in the future.2

2 The model developed here for Sri Lanka is expected to use in various policy simulation experiments

in the future

1.3 Outline of the Thesis

This thesis contains six chapters including the introductory chapter, bibliography and three appendices The outline of the thesis is as follows

Chapter 2 addresses the issue of sustainability of fiscal deficits We first extend the existing present value borrowing constraint model by incorporating rational expectations to allow for large information set to assess the discounted sum of expected future primary surpluses Then, we use the U.S federal government budgetary data over a long period of time to test the fiscal sustainability hypothesis Results emerge in favour of the sustainability hypothesis despite very divergent movements in deficits and debt series over the examined lengthy sample period

Chapter 3 is on the tax smoothing hypothesis We propose a new theoretical and empirical testing framework for the tax smoothing hypothesis We show that the optimal tax rate is linearly dependent on the permanent component of the government expenditure rate Based on this relationship and random walk in the tax rate, we propose that tax smoothing is valid if tax rate at time t and permanent government expenditure rate at time t-1 are cointegrated over time The proposition is put into test empirically using data of six OECD countries for the period 1950s to the present Results come into sight in favour of weak form of tax smoothing for all the cases

In Chapter 4, a medium scale macroeconometric model is developed for Sri Lanka to carry out policy simulation experiments Our model differs from existing models for Sri Lanka due to many reasons Theoretical consistency is maintained throughout the model Both aggregate demand and aggregate supply decisions are adequately modeled Since our prime objective is to examine the effect of fiscal deficit reduction measures, an extensive treatment is given to government budgetary

operations The model allows for short-term fluctuations of variables around their long-term relationship as it is estimated in an error correction format The model appears well-suited for policy simulation experiments as it performs very well in both static and dynamic simulation methods

Chapter 5 of the thesis examines the macroeconomic impact of reductions in various components of government expenditure The model developed in Chapter 4 is used for the policy simulation experiments The impact of government consumption, transfer payments and investment expenditure cuts on main economic aggregates and the fiscal deficit is examined by tracing-out short-term and long-term expenditure multipliers Results reveal that while falls in government consumption and transfer payment expenditures leave many macro variables unchanged, government investment expenditure cuts have significant negative impact on the economy As a result of consumption and transfer payments expenditure cuts the fiscal deficit falls significantly both in the short-run and in the long-run Despite the short-term negative impact, fiscal deficit rises in the long-run if government’s investment spending is lowered This long-term positive effect on fiscal deficit is due to an economic contraction

Chapter 6 highlights the main results of the thesis It also provides some policy implications and prospects for future research

Rest of the thesis gives Bibliography and appendices Appendices provide derivations of equations used in Chapter 2 and Chapter 3, cointegration and error correction methodology and details of construction of data for some variables used in Chapter 4

CHAPTER TWO

ON THE SUSTAINABILITY OF FISCAL DEFICITS:

THE UNITED STATES EXPERIENCE

2.1 Introduction

In many developed and developing countries fiscal balance has showed marked deteriorations during the past few decades leaving governments with persistent deficits Because governments mainly stick to a policy of bond-financing of deficits, large and persistent deficits cause sharp increases in the stock of public debt

An important issue of running continuous fiscal deficits and excessive accumulation

of debt is that how long a government could continue such a policy unchecked (Hamilton and Flavin (1986)) The issue, in general, concerns the sustainability of recurrent deficits

Large and persistent fiscal deficits may have many macroeconomic consequences In the Keynesian model, deficits are likely to have real effects and distributional impact by increasing current consumption and reducing future wealth (Barro (1974)) This is true under the assumption that an increase in government debt implies an increase in perceived household wealth It raises current consumption and thus, reduces capital accumulation and output growth On the other hand, public debt competed with private debt for available funds drives up interest rates crowding-out private investment Again this has a deleterious effect on the long-term economic growth (Seater (1993)) In the Diamond overlapping-generations model (see Diamond (1965)) where agents have finite lifetimes with no bequest motives, and taxes are non lump-sum, deficits have real effects as some of future tax burdens of a bond issue lie

on individuals who are not alive when the bond is issued Thus, it reduces wealth of some individuals and thus of the economy (see for example Modigliani (1961), Mundell (1971), Blinder and Solow (1973), Barsky, Mankiw, and Zeldes (1986), Bernheim (1987), Bernheim and Bagwell (1988)).3 Continuous deficits also involve departure from optimal policy of tax smoothing Under the tax smoothing hypothesis deficits are chosen optimally to minimize the present value of distortion costs of taxation (Barro (1979), Romer (2006, p.573)) Large deficits require the expected future tax rate to be higher than the current tax rate if the intertemporal budget constraint of the government is valid Thus, the deviation from tax smoothing means that the government creates an unnecessarily high distortion costs by imposing high tax rates to finance its excess spending

Provided that the economy is non-Ricardian, costs of deficit are obviously high if it is unsustainable.4 An unsustainable fiscal policy cannot continue to the indefinite future as it leads to a crisis A crisis laden fiscal policy may involve sharp contractions in the size of the government, thus, a large fall in aggregate demand This would ultimately lead to defaults in public liabilities which would in turn lower government spending and escalate economic recessions An example is the debt crisis

in early 1980s: A huge accumulation of debt as a result of large and recurrent fiscal deficits led many countries into debt crisis and defaults Subsequently, the defaulted countries endured severe economic hardships as creditors (private and official)

3 There is, however, no consensus that this assertion is always true In a Ricardian economy in which agents foresee that public debt has to be retired eventually and thus the government has to adjust expenditure accordingly, deficits have no real effects (see for example Barro (1974), Poterba and Summers (1987))

4 For a detailed discussion on the economic costs of sustainable and unsustainable fiscal deficits, see Romer (2001, pp 573-576, 2006, pp 603-607)

refrained from refinancing them Thus, what is important here for better macroeconomic management is to maintain the government budgetary operations within the limits of its intertemporal budget constraint

In this chapter, we examine the sustainability of the United States (U.S.) federal government’s fiscal policy The escalating fiscal deficit of the U.S federal government in recent decades has brought to surface the old fears of its sustainability Any perceived unsustainability of the fiscal deficit combined with the current account deficit which has reached staggering heights in recent years may severely affect the reserve currency status of the U.S dollar and may bring about a destabilizing effect

on the world economy Given this scenario it is worth examining how the past experience on fiscal operations shed light on the sustainability issue In a seminal paper, Hamilton and Flavin (1986) addressed this issue drawing evidence from the 1960-1984 period and reached a conclusion in support of the solvency of the U.S government The objective of our exercise here is to extend the Hamilton-Flavin methodology to incorporate rational expectations on future primary surpluses and examine the issue over a much longer time span that covers dynamically very different deficit episodes

The rest of the chapter is organized as follows A discussion on the methods of assessing fiscal sustainability is given in Section 2.2 In Section 2.3, we present the basic analytical framework of the present value borrowing constraint approach Section 2.4 discusses some limitations of existing models that have examined the issue of fiscal sustainability in the U.S and elsewhere within the framework of the present value borrowing constraint In Section 2.5, we use a rational expectations formulation to accommodate for non-stationary behaviours in the debt process and

present a more flexible model to test the present value borrowing constraint Issues of the sample period and data are given in Section 2.6 Section 2.7 highlights the salient features of the U.S debt and fiscal balance series that span over 75 years In Section 2.8, we test the present value borrowing constraint using data over this long time span that has recorded very different dynamics of the debt process It should be noted that our model captures all the important turning points of the debt series very well Overall, the results emerge in support of the solvency, perhaps the super-solvency, of the U.S federal government in the long run

2.2 Methods of Assessing Fiscal Sustainability

There are two broad conceptual approaches to assessing fiscal sustainability: the accounting approach and the present value borrowing constraint approach.5 In addition to these main approaches, this section briefly discusses the implications of the Ricardian equivalence theorem, bona-fide fiscal policies, strategic default models

of government debt, Modigliani-Miller type results for fiscal policy on the issue of fiscal sustainability

The accounting approach uses few indicators, mainly debt-income ratio (or the debt ratio), to examine how far the present fiscal policy departs from a sustainable one Consequently, a primary deficit which is defined as the excess spending excluding interest payments on existing debt over revenue (including money financing or seignorage) is treated as sustainable if it generates a constant debt ratio.6 Buiter (1985) notes that a sustainable fiscal policy needs to maintain the debt ratio at its current level For Pasinetti (1998) and Goldstein (2003), a sustainable fiscal policy is the one

5 Cuddington (1997) and Chalk and Hemming (2000) provide excellent surveys of the literature

6 Seignorage is defined as government revenue from printing money (Romer (2006, p.538))

that generates a stable debt ratio overtime Blanchard et al (1990) note that fiscal policy is sustainable if the debt ratio eventually converges to its initial level As for the accounting approach, sustainability is essentially about whether a government is heading towards an excessive accumulation of debt under the current policy However, this method is often questioned on the lack of a proper threshold criterion for the debt ratio in evaluating the fiscal sustainability In particular, this threshold value of debt ratio should depend on the fundamentals and the requirements of the economy

In contrast, the present value borrowing constraint approach provides a solid theoretical base for testing the sustainability hypothesis It implies that fiscal policy is sustainable when it is expected to generate sufficient net revenues in the future to repay the accumulated debt and interest expenses That is, fiscal policy is sustainable

if it can be maintained into the indefinite future without leading the government into insolvency Public sector is solvent when the present discounted value of future primary surpluses is at least equal to the outstanding stock of debt On the contrary, the policy is said to be unsustainable when the government will forever finance its interest payments by issuing new debts In the case of individuals’ budget constraint, Ponzi schemes are ruled out as they are infeasible: No one would be willing to lend a person who is trying to roll over debt continually.7 Similarly, fiscal sustainability requires deficit/surplus policies to be subject to its intertemporal budget constraint As Ponzi schemes are not possible in a dynamically efficient economy where the real rate

of interest is higher than the output growth rate, this implies that the present deficits should be offset by surpluses in the future.8 Though there may be short-term

deviations from it, the constraint cannot be breached in the long run Empirical evidence on the violation of the present value constraint indicates the unsustainability

of the policy.9 Hamilton and Flavin (1986) initially use this method to evaluate fiscal sustainability

In the Ricardian equivalence view if future tax liabilities implicit in deficit financing are accurately foreseen, the behaviour of the agents will be exactly the same

as if the budget is balanced continually Since deficits have to be eventually retired, the policy is sustainable so long as it does not violate non-negativity of individuals’ consumption (see Barro (1974))

Assuming bonds created by the government are potential stores of value and they are used by households in paying taxes and by the government in making transfers in a complete and transaction costless market, Balasko and Shell (1981) show that Ponzi-type schemes are possible to an extent in an infinite-horizon overlapping-generations model Treating all forms of government bonds as money, they find that the policy is bona-fide if the price of money is positive in all periods If long-run interest rates exceed the long-run output growth rate, the bona-fide monetary (in fact, fiscal) policies entail long-run money stock of zero Balasko and Shell (1981) noted that a policy that allows money supply to grow faster than the real growth rate asymptotically is not Pareto optimal Though it is not straightforward, this implies that

if debt grows at a higher rate than the output growth rate the policy is not sustainable

as Ponzi schemes are feasible in such situation

9 A detailed discussion on the intertemporal budget constraint and the present value borrowing constraint is given in the next section

In strategic default models of government (foreign) debt, in which the level of debt may be determined by the borrower’s demand for credit or by a credit ceiling imposed by the lender, the government may have increased incentives to default when the debt stock is large If the penalty for defaults is just the expulsion from foreign

capital market, a country can borrow on an uninterrupted basis until period t with no

further intention of borrowing thereafter (see for example Eaton and Gersovitz (1981), Bulow and Rogoff (1989)) Thus, the issue of debt sustainability depends on incentives for the government not to repudiate its debts

Wallace (1981) shows that there is a Modigliani-Miller type result for fiscal and monetary policies.10 If there is an equilibrium with certain properties for one path

of portfolio for the government, then the above equilibrium is an equilibrium for a large class of portfolios of the government provided that only lump-sum taxes are adjusted to hold fiscal policy constant That is, alternative paths of government portfolio consistent with a single path of fiscal policy can be irrelevant Irrelevance here means that both the equilibrium consumption allocation and the path of price level are independent of the path of government portfolio Though this thesis has no direct implications on fiscal sustainability, the indeterminacy on government’s portfolios could make the issue of sustainability unclear

With no proper criteria in the accounting approach and no direct implication

on the sustainability in other approaches such as the Ricardian view, bona-fide fiscal policies, strategic default models and Modigliani-Miller type models, we reckon that the present value borrowing constraint approach provides a solid theoretical base for fiscal sustainability that could be tested appealing to past budgetary policies of the

10 The Modigliani-Miller theorem shows that alternative corporate liability structures are irrelevant (see Modigliani and Miller (1958))

government Obviously fiscal sustainability is a concern on past deficits/surpluses policies of the government And the present value borrowing constraint approach enables testing the issue with the application of advanced econometric knowledge and techniques For these reasons, we follow and extend the present value borrowing constraint approach in testing the fiscal sustainability of the U.S federal government

2.3 The Analytical Framework

Following the work of Hamilton and Flavin (1986) we assume that the government issues only one kind of bond , the aggregate debt, and its marginal cost is given by r , the real interest rate We invoke rational expectations here to set

the future real interest rate to its conditional mean (see also Hansen, Roberts, and Sargent (1991), Roberts (1991)) Although the conditional expectation of real interest rate,

t D

1

( t t)

E r I+ , is a random variable because of the changing information set I , by t

taking iterated expectations we can write E r t E I (E r t I t ) r

the mean of the conditional expectations of real interest rate is a constant This condition is empirically valid if the rate of real interest is a stationary (I(0)) process Since a unit root in the U.S real interest rate is rejected, it is reasonable to assume that the unconditional mean of real interest rate is constant at its period average.11 We further assume that the government’s borrowing begins with a given initial condition

at time t and ends in the period t N+ −1 where N is an integer greater than one

Agents who lend the government in each period believe that the government will run

11 We found that ADF unit root test reject a unit root in the U.S real interest rate at the 1% level of significance The ADF unit root test static with a constant and two period lagged effect in the regression is -4.942 for the period 1932-2004 The ADF critical value at the 1% significance level is - 3.52 See Table 2.3

sufficient primary surpluses in the future, up to t+N, to offset its initial debt The

government’s instantaneous budget constraint takes the following form:

S = − is the primary surplus, is government revenue and is

non-interest government expenditure Writing (2.1) as

t

t G

i N

t t N

In (2.4), the current debt stock is equal to the expected present-value of the debt stock

in the limit (A t ) plus the discounted sum of the expected future primary surpluses

( *) If the no-Ponzi-game (NPG) condition holds, i.e.,

t

then the debt stock at time t must be matched by the present-value of expected future

primary surpluses If the NPG condition holds in strict equality, , the

Ponzi schemes are possible in an overlapping-generations model in an infinite horizon economy In such environments, there would be some imposed reallocation of resources which make no individual worse-off while making at least one individual better-off (Samuelson (1958)) Shell (1971) notes that this result is an outcome of the

“double infinity” of traders and commodities in the model In such environments making the future generations to pay tax liabilities, the government can issue debt at one period and roll it over forever (see Shell (1971), Romer (2006, p.563)) For the feasibility of this scheme, it is required that the real rate of interest to be less than the growth rate of the economy Since governments issue debt to repay the principal and interest on the initial debt, the debt stock grows at the rate of real interest As the real interest rate is less than the output growth rate, this policy yields a consistently falling debt ratio and thus there is no fear of continuing the policy However, such a policy violates the conventional budget constraint as the value of discounted debt is a positive constant in the limit It is also argued that when uncertainty is present in decision making, Ponzi schemes may be feasible even if the economy is dynamically efficient This is because with uncertainty the realized rate of return on government debt may be less than the economy’s growth rate (see Bohn (1995), Ball, Elmendorf, and Mankiw (1998), Blanchard and Weil (2001))

In reality the government is also unable to run Ponzi schemes as of individuals because of dynamically efficient nature of the economy Suppose the limit value of the present discounted debt is strictly positive (A t > 0) If there a finite number of

agents in the economy, A t > 0 means that the present value of at least one agent’s wealth in the limit is strictly positive This is, however, not an equilibrium point as the agent can increase his utility by slightly increasing his spending As Ponzi schemes

are ruled out in dynamically efficient economies, the government must satisfy the traditional present-value borrowing constraint (see O’Connell and Zeldes (1988), Romer (2006, p.564))

where A0 =limN→∞ E t[ 1( +r)−N D N] and (1+r) t is deterministic bubble term

If A is treated as a constant, then the present value borrowing constraint holds if 0

Hamilton and Flavin tested this condition in three ways The first test resorts

to examining the stationarity of

A ≤

t

D and in (2.5) Hamilton and Flavin argued that

stationarity of both these series necessarily implies that

*

t S

0

A = 0 They found that D t

and to be stationary over the period 1960-1984 and concluded that S t A = 0 Here 0

they relied on the condition that stationarity of implies stationarity of In the second and third tests Hamilton and Flavin assumed adaptive and partial-rational expectations respectively and regressed

t

t S

t

D on (1+r)t and current and lagged values

of S t The estimated A 0 coefficient turned out to be negative and insignificantly different from zero and confirmed their former conclusion

Though it has many applications (see Smith and Zin (1991) for an application

to Canadian data), several questions arise in relation to the Hamilton-Flavin testing procedure First, the stationarity of the debt series depends on the value of the real

interest rate If r >0 (Hamilton and Flavin use r = 0.0112) D becomes a non- t

stationary process (see (2.1)) The fact that the observed D was stationary over 1960- t

1984 indicates that r t was negative for sufficient number of periods Second, when the discount factor is close to unity (0.9889 in the Hamilton-Flavin study), the process becomes virtually a unit root process unless the sample size is extremely large (perhaps 500 years or more)

*

t S

13 So the use of current and lagged values of stationary process does not capture the dynamics of the process Third, if is stationary the correlation between and (1+r)

Wilcox (1989) offers a much simpler method for testing the present value

borrowing constraint First, he relaxes the assumption of fixed r, and defines a

variable discount factor by =∏− ( + )

series A′ can be non-stochastic if and only if expectations of the limit value of debt t

are the same in each period A non-stochastic A′ implies that the expected t D′ in the t

limit does not change upon the arrival of new information If expectations of D′ in t

the limit change with new information, A′ becomes stochastic Then, from (2.6) it is t

obvious that the path of A′ is determined by the path of t D′ : if t D′ is stationary, t A′ is t

constant, and if D′ is non-stationary, then, t A′ is not a constant Therefore, a zero- t

mean stationary D′ series satisfies the present value borrowing constraint Using the t

Hamilton-Flavin sample, Wilcox computed D′ series and found it to be non-stationary t

with a positive unconditional mean This implies that the discounted debt series grows over time and the limit value of the debt increases continuously Wilcox, therefore, concluded that the U.S federal fiscal policy was unsustainable

Corsetti and Rubini (1991) use Wilcox’s approach to test the fiscal sustainability by allowing a positive drift and a time trend in D′ series In a sample of t

OECD countries, they find mix results over the fiscal sustainability However, their results on the U.S confirm the Wilcox finding Buiter and Patel (1992), Baglioni and Cherubini (1991) and Gerson and Nellor (1997), using Wilcox’s method, test the sustainability of fiscal policy in a developing country context applying it to India, Turkey and the Philippines respectively Uctum and Wickens (2000) used a similar method to analyze the U.K fiscal data

Wilcox’s procedure is attractively simple However, it raises several concerns First, Wilcox’s method is to make a backward formulation A government standing at

time t discounts its debt D t to some initial year t=0 using an observed q t In reality a government faced with a debt stock at time t does not discount this back to a year

in the distant past to assess whether the discounted in the limit is zero or not The government has to look into the future and formulate policies to deal with the debt

t D

t D

stock, It should be noted that although (2.6), rather the expression in Footnote 13,

is derived from the accounting identity (2.1), as a behavioural relationship it is qualitatively different from (2.4) An important question to ask is, is it meaningful to derive the expected present-value of future surpluses or the discounted values of future debts using variable discount rates which are unknown? Governments and economic agents are more likely to engage in a scenario analysis by computing several present-values of projected surpluses by using a range of discount rates that are set to decline over time in a systematic way If that were the case, the assumption

of constant conditional expectation on interest rate is plausible

t

D

Second, Wilcox focused on A′ of (2.6) and argued that t series should have

a zero-mean in the limit for the present-value constraint to hold This limit value approaching zero is true by construction Since the discount factor has to approach zero in the limit, should approach zero unless the debt series itself explodes without limit

t D′

t D′

15 But what is important to note is that the expected limit value in (2.6)

is zero regardless of the I(0) or I(1) nature of the series For instance, consider the case where is I(1) and hence

t D t

( ) [ ] ( )

limN→∞ E D t ⎡ +aN+ i N= u N i− ⎤/ 1+r N =limN→∞ E D t +aN / 1+r N =0

where the initial value, D 0, is assumed given and r >0

For fiscal sustainability what is important is how fast the discounted debt series approaches zero Given the observed interest rates, it would be extremely difficult to construct an A′ series that would converge to zero within the time span of t

the recorded data series For instance, a huge debt burden over 25 years may be sufficient to cripple a government even though the discounted debt becomes zero after

200 years Another point to note is that the constructed discounted debt series involves a scaling effect The level of debt is likely to increase with the size of the budget that in turn increases with the size of the economy Without removing this scale effect it would be difficult to see the series D′ converging to zero Wilcox’s t

method may be better implemented if debt is taken as a ratio of GNP to remove the scaling effect

Trehan and Walsh (1988, 1991), Hakkio and Rush (1991) and Haug (1991) focused on a condition that revenue minus expenditure inclusive of interest payment must be stationary for the present value borrowing constraint to hold This requires that the revenue and expenditure series to be cointegrated with a vector (1 -1) The intuition here is that the budget is balanced in the long-run despite short-term deficits

or surpluses Tanner and Liu (1994), Quintos (1995), Haug (1995), Martin (2000), Cipollini (2001) and Jha and Sharma (2004) extended this method allowing for structural break(s) Nonetheless, the cointegration between revenue and expenditure series with a vector (1 -1) is embedded in the stationarity of the surplus series Bohn (1995, 2006) argued that the deterministic intertemporal budget constraint is

inappropriate and tests on stationarity and cointegration restrictions are invalid when the interest rate is stochastic.16 Hansen, Roberts, and Sargent (1991) and Roberts (1991), however, show that the validity of the deterministic intertemporal budget constraint only requires the conditional expectation of the interest rate to be constant

over time The interest rate could be time varying ex post Ahmed and Rogers (1995)

argue that under some plausible assumptions cointegration tests on the present value borrowing constraint are valid even in a stochastic environment

Some recent studies employ Markov-switching approach to describe the

non-linearities present in the discounted D t series (see Bajo-Rubio, Diaz-Roldand, and Esteve (2004), Davig (2005)) Nonetheless, this univariate regime-switching method does not account for the non-linearity in the budget variables The trending patterns of the discounted debt series could be a result of non-stationarity in fundamental

variables such as income, G 0 , T and other relevant variables It is also important to

emphasize that structural breaks and non-linearities in a single series do not necessarily amount to a breakdown in a causal relationship (see Hoover (2001) for an excellent exposition of this point)

In another line of argument, Kremers (1989) argued that if fiscal policy yields

a stock of debt in real terms that grows asymptotically at an average rate smaller than the interest rate, the policy operates within the present value borrowing constraint

since the present-value of debt converges to zero as N becomes large.17 However, government borrowing would not be restricted by the present value borrowing

16 Bohn (1995) and Ahmed and Rogers (1995) test the fiscal sustainability with uncertain marginal cost

of borrowings In place of the real rate of interest, they use marginal rate of substitution of consumption

17 Note that the paths of debt and primary balance associated with this policy may involve taxes that grow faster than the taxing capacity of the economy This violates the government’s collateral for future borrowings Therefore, it is reasonable to assume that taxes are limited to a fraction of income

constraint if the equilibrium is dynamically inefficient.18 If there is no sufficient evidence on the dynamic inefficiency of the U.S economy (see Abel et al (1989)), Ponzi financing of fiscal deficits is ruled out in the long-run For McCallum (1984), fiscal sustainability does not reject the government’s ability to run a permanent deficit inclusive of interest payments A permanent deficit exclusive of interest payments, however, violates the present value borrowing constraint and hence is not sustainable

in the long run.19 Bohn (1998a) suggested that fiscal policy is sustainable if the primary surplus positively responds to the changes in debt-income ratio as it provides direct evidence for corrective fiscal policy actions Based on this line of argument he concluded that the U.S fiscal policy has historically been sustainable Since changes

in spending or taxes are necessary if the current fiscal policy does not satisfy the intertemporal budget constraint, Auerbach (1997) proposes a measure of the size of the expected fiscal imbalance Given the projected paths of government revenue, income, spending and interest rates Auerbach criterion quantifies the required tax hike (or spending cut) in order for the intertemporal budget constraint to be satisfied Applying this framework for the U.S federal budget, Auerbach et al (2003) conclude that the current U.S fiscal policies are quite far from satisfying the intertemporal budget constraint Thus, the current U.S fiscal policy requires significant changes to achieve fiscal sustainability In a political-economy model, Velasco (2000) shows that due to the common pool nature of government resources fiscal deficits emerge

regardless of the intertemporal budget balance Therefore, government debts tend to

be excessively high in the long-run

Given the implications of these various approaches on fiscal sustainability, an advantage of focusing on the full equation in (2.6), which is derived from the intertemporal budget constraint of the government, is that it may remain intact by structural breaks and regime switches It also depends on the fundamental budget variables and thus takes care of the scaling effect on the level of debt In this context, the Hamilton-Flavin regression may perform better over a longer time span since it can accommodate non-stationary series in a cointegrating framework

2.5 Methodology of the Present Study

We examine whether the present value borrowing constraint holds or not directly as in the Hamilton-Flavin model (2.6) However, the discussion in the previous section highlights the need of a more flexible model structure to account for non-stationary movements of the debt series We have already noted that when the discount rate is close to unity the expected present value of primary surplus, , may behave like a unit root process in observed samples If also shares a similar behavior it is still possible for the present value borrowing constraint to hold if and

cointegrate (locally) such that

t

D

*

t S t

Since is unobserved, we adapt the rational expectations formulation

developed by Hansen and Sargent (1980) to relate to observed variables (see also Sargent (1978), Wallis (1980) and Campbell and Shiller (1987)) For this, we assume

*

t S

*

t S

that the policymakers form expectations on the discounted sum of future surpluses

using all the relevant information available to them at time t as given below

where ρ = 11( +r), Z t =a X′ t , a is an (n×1) vector of constants and X is an (n×1)

vector of relevant informational variables known both to the government and the

public, and w t is an unsystematic informational variable only known to the government with the property E w t t k+ = The information set X0 t may include both stationary and non-stationary variables that are implicit and explicit in future budget

deficits/surpluses Z t , a linear combination of variables in the vector X, is stationary

but it shares the same near unit-root behavior that is likely to possess in observed samples Assuming that

*

t S t

Z has the following infinite-order moving average

Since the non-linear parameter structure in this formulation is not in our interest,

can be written as a linear function of current and past Z

*

t S

t series:

1 0

p k

t

ε is a well behaved disturbance term.21

Model (2.10) entails a couple of advantages over the previous formulations First, it is grounded on a larger information set that policymakers and economic

agents are likely to use at time t to make projections about future surpluses This large information set would also enable endogenous effects of variables determining S t and

D t series Second, without incorporating structural breaks or regime-shifts artificially

it can accommodate non-stationary movements of D t that do not violate the present

value borrowing constraint Non-stationary movements of D t could be well explained

by (2.10) as it provides a cointegration framework between D t and other fundamental budget variables However, if there are breaks or regime shifts that disturb estimated parameters significantly one has to incorporate these factors to the proposed model for better results

21 A description of variables in vector X and coefficients in vector βis given in Section 2.8 following the data adjustment procedure given in Table 2.2

2.6 Sample and Data

In this study, we use the U.S federal budgetary data (annual) for the period 1929-2004 because it covers very divergent deficit episodes and trends in the debt series (see Figure 2.1) Data on interest bearing gross public debt, expenditure, interest payments on the existing debt, base money stock, stocks of foreign liquid assets are obtained from the Federal Reserve Archival System for Economic Research (FRASER) Annual observations of GDP and implicit GDP deflator are from the National Income and Product Accounts (NIPA) Data of the unemployment rate are from the U.S Bureau of Labour Statistics Real values of relevant variables are given

in 2000 prices

2.7 Trends in Deficit and Debt of the U.S Federal Government

This section discusses the salient features of the U.S federal fiscal operations from 1930 to 2004 Table 1 provides details of revenue, spending, interest payments

on past debt, overall and primary balance and interest bearing gross public debt as a percent of GDP We construct net stock of interest bearing public debt stock.22 The primary balance has been adjusted in line with the net stock of public debt We made these adjustments to the gross debt and officially reported surplus series following the comments made by Eisner and Peiper (1984) and Eisner (1989) They noted that official measures of federal debt and budget deficits/surpluses are misleading because

22 However, many studies use market value rather than par value of debt Market value reflects the amount of debt that has to be financed by taxes if debt were to be paid off (in part) in a given period Seater (1981), Butkewicz (1983), Cox and Hirschhorn (1983), Cox (1985) are some painstaking attempts to calculate market values of the U.S government debt However, to quote Hamilton and Flavin (1986) “[a]t the end of fiscal year 1974, outstanding government debt [of the U.S.] was trading

at a market value of only 95 percent of its par value” Thus, we reckon that the difference between market and par values of debt is trivial Because of the difficulty in computing market value, we stick

to the use of par values in the analysis Nonetheless, this will not have significant impact on the main analysis and conclusions of this chapter

they do not account for other relevant accounts of the government Gross public debt figures ignore the accumulation of financial and real assets which have contributed to

an increase in government net-worth Moreover, the official measures of surplus/deficit need to account for the changes in government financial assets to be consistent with the changes in the real net debt

Table 2.1 Fiscal Summary of the U.S Federal Government (as a % of GDP)

Year Revenue Spending Interest Fiscal deficit / surplus Gross

payment a Overall Primary b debt c

Source: Author’s calculation based on FRASER and NIPA data

Notes: (a) interest payments on government debt; (b) revenue minus non-interest expenditure; (c) interest bearing gross public debt

The adjustment method is illustrated in Table 2.1 We subtract the stock of high powered money and the stock of government foreign financial assets from gross debt to obtain net stock of debt The government’s foreign financial assets here include the stocks of gold, SDR, reserves at the IMF and convertible foreign currencies The high powered money and foreign financial assets are used as an approximation to the financial assets of the government The primary balance is adjusted by accounting for changes in high powered money and changes in foreign

financial assets The adjustments made here, to some extent, are similar to those made

by Hamilton and Flavin (1986)

Table 2.2 Computation of Net Debt and Adjusted Primary Balance-2004

and convertible foreign currencies (FA) - 83.5 (=) Net stock of debt (at current prices) 6527.2 ( ) by implicit GDP price deflator ÷ ÷ 1.082 (=) Real net stock of debt (at 2000 prices) 6031.4 Adjusted primary balance

Note: The sum of H and FA (foreign liquid assets) is used as a proxy for the financial assets

of the government

Source: Authors’ calculation based on FRASER and NIPA data

Figure 2.1 plots the both unadjusted and adjusted data series It is worth reporting some observations from these data series Figure 2.1b shows that though there is a level difference, the time paths of both gross and net debt are quite similar Figure 2.1(a) shows that though the adjusted primary balance is often higher than the unadjusted one, they also follow the same pattern over time Between 1929 and 1942 the net debt stock was quite small (negative in some periods) (see Figure 2.1(b)) As a result of wartime (World War II) high deficits, the debt stock rose to a staggering 124